I don't think anyone has answered my question of how
is this different than the Muller-Plathe method. For
viscosity the MP method is almost always better than
the traditional NEMD way of computing it. I'm guessing
the same is true of MP for thermal conductivity versus
the GK method. MP will converge much more quickly
than GK b/c the induced temp profile in MP is stable, whereas
the induced heat flux in GK is a quantity with large fluctuations.
In viscosity it is the off-diagonal pressure component that fluctuates widely
and MP avoids computing it at all.
I also disagree with the statement below that GK can easily
be implemented for many-body potentials. It would require
inserting new code in every such potential in LAMMPS (AIREBO, Tersoff,
MEAM, ReaxFF, etc). By contrast the MP method just
works with any potential without the need to alter the potential at all.
It also just works with long-range Coulombics.
Are there other downsides to MP for thermal K that I am missing?
Steve